So, after three blog posts, what is the moral of this Blue Ridge Numerics’ “CFD Comparison Chart” saga?
Well, clearly one should not believe every “CFD Comparison Chart”! – or even the many “charts” that one vendor has put out. Rather, the story suggests that people who are thinking of buying a CFD tool for the first time might be well advised to be wary of what vendors claim about their own software relative to or as well as other company’s products. I’d also be very wary of vendors who use simple CFD demo models to show the software for your application.
For me, the lesson of the story is this: Do your own research! Caveat Emptor (“let the buyer beware” in Latin) has long been a principle in property law, and well worth keeping in mind whenever you’re buying anything you want to provide lasting quality.
On a more positive note, here’s some more advice from my 20+ years of experience in CFD that should help if you’re thinking of buying a CFD code for the first time:
Dr. J’s 8 Steps to Successful CFD Acquisition
- First think carefully about what features you need the CFD tool to have for your particular set of applications. What capabilities do you need? Do you need to simulate compressible flows, combustion, cavitation in liquids, particulates etc.?
- Next ask yourself who will use the tool. Is it someone with a background in CFD or a mechanical engineer who understands the physics, but has no background in CFD?
- What’s the environment in which the tool will be used? Is it a research environment where result accuracy is of paramount importance, or a product design environment where the impact of geometry changes need to be assessed overnight?
- Get the vendor to give you the technical information you need. Ask to see relevant CFD benchmarks/validations for the applications you are considering and testimonials from customers in the same industry as you, or dealing with similar applications. Ask to see industrial-scale demos (as some CFD codes struggle with large numbers of objects - see below for some other specific technical points to watch out for)
- Find out what licensing model the vendor uses. Be wary of vendors who only sell perpetual licenses, particularly through resellers, as it reduces the incentive for the vendor to provide good quality support. Not all companies provide the 5-Star support for which Mentor is famous!
- Go on a training course. Yes do! The benefit is two-fold. First you get trained by people who are expert in using the software, so you’ll get the money back in improved productivity pretty quickly. Second, agree up front that you can take along something that’s representative of what you want to model so you can see how easy it is to set up and get to a first result.
- Consider taking an annual license for the first year – it reduces the financial risk and helps ensure the vendor is motivated to give you the best support when you need it most.
- If you’re planning to use the software in a business-critical area like product design then look at the vendor’s business. How long have they been in business? How large are they? How stable are they? Are they publicly traded? Large, publicly traded companies tend to be more resilient …
Technically watch out for:
- Convergence: Even some relatively simple flow simulations need to be well-converged before the results can be believed; for example, to accurately predict the development of a laminar pipe flow it’s necessary to correctly capture the very small radial velocity components in the flow. Not all CFD codes converge well. Finite-element based codes are notoriously difficult to converge, which partly explains why the vast majority of both commercial and in-house CFD codes are finite-volume based today. Get the vendor to show you the convergence history and proof your application is converged. This applies to point 4 above.
- Natural Convection: Some codes just don’t handle natural convection cases well, despite the importance of natural convection effects in many real world fluid flow and heat transfer applications. Ask to see a natural convection case; again with the convergence history and proof it’s converged.
- Mass Balances: In steady-state cases, checking the mass balance across the system is a good way to check for convergence in a CFD code. You should be able to see that there’s a near zero mass balance difference between the inflows and outflow. The generally accepted level of error is 1% of the flow through the system, or less, taken as the sum of the absolute value of the error in every cell. Simulation results with a 20% error say are simply meaningless, as the true physical behaviour may be very different!
- Meshing: Some meshing approaches require a lot of manual effort to produce a good quality mesh. Some mesh topologies that are easier to automate, such as tetrahedral cells (i.e. a tet mesh), offer the lowest solution accuracy, and are not well-suited to resolving long narrow channels, as the number of cells required along the channel quickly makes the case intractable and artificial diffusion may prevail. Similar comments apply to any slender geometry that has a high aspect ratio. Be wary if the vendor ignores small but important geometric features, or uses a less challenging geometry to demonstrate their code.
If you’re working in a fast-paced design environment and need a robust tool that design engineers can use I would encourage you to check out the breadth of applications that FloEFD covers despite what some “CFD Comparison Charts” might say about it, which includes complex physical phenomena like non-Newtonian fluids, compressible flow, combustion and cavitation. FloEFD is CAD-embedded, allowing you to evaluate the product’s performance concurrently with changes to the design – something we term “Concurrent CFD”. To get a feel for how Concurrent CFD is different to Traditional CFD approaches and Upfront CFD, take a look at these blog posts…
Concurrent CFD Explained (Part I)
Concurrent CFD Explained (Part II)
Concurrent CFD Explained (Part III)
Concurrent CFD Explained (Part IV)
And do remember that not all CFD Comparison Charts are equal…or right!
Dr. J, Hampton Court